RESEARCH & IDEAS
Risky Business with Structured
Finance
Published: January 20, 2009
Author: Julia Hanna
How did the process of securitization
transform trillions of dollars of risky assets into
securities that many considered to be a safe
bet? HBS professors Joshua D. Coval and Erik
Stafford, with Princeton colleague Jakub
Jurek, authors of a new paper, have ideas. Key
concepts include:
• Over the past decade, risks have been
repackaged to create triple-A-rated
securities.
• Even modest imprecision in estimating
underlying risks is magnified
disproportionately when securities are
pooled and tranched, as shown in a
modeling exercise.
• Ratings of structured finance products,
which make no distinction between the
different sources of default risk, are
particularly useless for determining prices
and fair rates of compensation for these
risks.
• Going forward, it would be best to eliminate
any sanction of ratings as a guide to
investment policy and capital requirements.
• It is important to focus on measuring and
judging the system's aggregate amount of
leverage and to understand the exposures
that financial institutions actually have.
In the wake of the financial crisis, many
once-esoteric investment terms have become a
familiar part of our vocabulary. The role of
structured finance securities such as
collateralized debt obligations (CDOs), for
example, and the part played by ratings
agencies in legitimizing these products, has
become all too clear. The pooling and
repackaging of economic assets such as loans,
bonds, and mortgages resulted in enormous
yields for many investors—until, one day, they
didn't.
"The Economics of Structured Finance," a
paper [PDF] forthcoming in the Journal of
Economic Perspectives, offers a close
examination and clear explanation of how the
process of securitization transformed trillions of
dollars of risky assets into securities that many
considered to be a safe bet. Authored by HBS
professors Joshua D. Coval and Erik Stafford,
with Jakub Jurek (HBS PhDBE '08), an
COPYRIGHT 2007 PRESIDENT AND FELLOWS OF HARVARD COLLEGE
assistant professor at Princeton University, the
paper analyzes the difficulties of rating
structured finance assets and the perils of
relying on ratings to determine prices.
"We began studying the corporate bond
CDO market roughly three years ago and
reached the conclusion that it appeared to be
severely mispriced," Coval remarks. "But we
didn't have any sense that it would all come
crashing down the way it did—we just thought
that the mispricing would be corrected
gradually."
Stafford adds that neither he nor Coval had
done much work in the area of credit
markets—yet these markets were growing at
such an unprecedented rate that it piqued their
curiosity. "People didn't explain anything about
CDOs in a compelling way," he says. "They
said, 'Oh, they're slicing and dicing.' Or they
resorted to mathematical descriptions that were
lacking any economic explanation as to why
this product was so superior to other channels
of financing. We also started hearing the phrase,
'These structures create yield out of thin air.' It
implies magic, or a violation of a very standard
notion in finance, a Modigliani-Miller
proposition that the way you finance something
is irrelevant to the value of the assets."
In their paper, the authors recount how, over
the past decade, risks have been repackaged to
create triple-A-rated securities. By mid-2007,
they write, 37,000 structured finance issues in
the United States had earned this top mark. In
2006, Moody's reported that 44 percent of its
revenues came from rating structured finance
products, versus the 32 percent of revenues
generated from its traditional business of rating
corporate bonds.
A simulation is "still a
model and it's not exact,
which is easy to forget."
-Erik Stafford
Yet the paper presents a modeling exercise
(using the same computer tools employed by
the ratings agencies) that demonstrates that the
challenge of rating structured products lies in
their extreme sensitivity to estimation
errors—that even modest imprecision in
estimating underlying risks is magnified
disproportionately when securities are pooled
and tranched. Their simulation uses pools
consisting of 100 bonds with a five-year default
probability of 5 percent and a recovery rate of
50 percent of face value conditional on default.
Within each pool, the exercise creates a
capital structure consisted of junior, mezzanine,
and senior tranches. The junior tranche absorbs
losses from the pool until the portfolio loss
exceeds 6 percent, at which point it becomes
worthless. The mezzanine tranche begins to
absorb losses at that point, continuing to do so
until the portfolio loss reaches 12 percent, with
the senior tranche absorbing losses in excess of
12 percent. The scholars also run a simulation
that constructs a "CDO²" by further dividing the
mezzanine tranches, noting that due to the
practice of subdividing large pools of
residential mortgages, many CDOs of
mortgage-backed securities were essentially
CDO²s. That, coupled with the increase in
subprime mortgages—from $96.8 billion in
1996 to approximately $600 billion in
2006—created a recipe for economic disaster.
The exercise clearly shows how the
sensitivity of tranches to error in the estimate of
default probability is determined by their
seniority. An increase in the default probability
from 5 percent to 10 percent results in a 55
percent decline in the expected payoff for the
junior tranche, an 8 percent decline for the
mezzanine tranche, and a .01 percent decline for
the senior tranche. The effect is amplified in the
CDO², with the value of the junior and
mezzanine tranches falling quickly toward zero
and the value of the senior tranche declining
substantially as default probabilities rise. Earlier
in their paper, however, the researchers
illustrate the difficulty of estimating the default
rate and assigning investment ratings of
comparatively straightforward corporate bonds
by citing 10-year historical data from Fitch on
their default probability. Within the 10
investment grade rating categories of AAA to
BBB-, the annualized default rate only varies
between .02 and .75 percent, leaving little
margin for error. The speculative grade range
(from BB+ to C), meanwhile, has a default rate
ranging from 1.07 percent to 29.96 percent.
1

Yet the ratings agencies aren't entirely at
fault. "You can blame them a bit for going
along with things and for overconfidence in
their abilities," says Coval. "They were trying to
rate these securities to the best of their abilities,
and Wall Street kept telling them to rate more.
They had a sense they were doing something
wrong, but they were trying to improve their
models as quickly as they could. I don't think
they were deliberately manipulating the models
to make them especially biased or imprecise."
The computer simulations used to create the
ratings are a standard tool, he says, adding that
many would also describe it as a dangerous one
because they create a false sense of confidence
in what are essentially estimates.
"It's impossible to use a pencil and paper to
characterize the interaction of a thousand
different securities with different recovery rates
and different business models," Stafford
comments. "With a simulation you can get close
to the right answer, but it's still a model and it's
not exact, which is easy to forget. Often the
people who are attracted to this approach are
mathematicians, not economists, so they might
not appreciate the underlying nature of some of
the things they're describing." The paper notes
that one neglected feature of the securities
produced by structured finance is that it
substitutes risks that are largely diversifiable for
COPYRIGHT 2007 PRESIDENT AND FELLOWS OF HARVARD COLLEGE
HARVARD BUSINESS SCHOOL | WORKING KNOWLEDGE | HBSWK.HBS.EDU
risks that are highly systematic, or linked to
events in the economy. As a result, ratings of
structured finance products, which make no
distinction between the different sources of
default risk, are particularly useless for
determining prices and fair rates of
compensation for these risks.
The paper also raises larger questions.
"There has been a significant change in the
price of these securities," says Coval. "How
much of the repricing that's occurred was an
elimination of mispricing that was prevalent
prior to the crisis? How much of it was a
spillover from the mortgage markets to markets
that were fairly priced? And if there was a high
degree of mispricing before the crisis, what
impact did that have for the level of credit that
was extended in the economy? How much of
the $9 trillion of assets that entered these
structures wouldn't have been originated in the
first place if you hadn't been able to put them in
these structures and 'create yield out of thin air'?
"That, to me, is probably the most important
question as we think about how deep this crisis
is going to go," Coval continues. "If the
economy over the last 10 years had, say, a
trillion dollars or more of credit extended than
was appropriate, given the risks that were being
borne, then we now need to go through a
significant adjustment."
The inevitable question, of course, is what
to change going forward. Coval advocates
eliminating any sanction of ratings as a guide to
investment policy and capital requirements.
"Without that, there's no cover for that investor
who says, 'I don't understand this product, but
it's rated AAA, and my board will accept that as
a good enough justification to hold it.' Over
time, we've relied on these ratings and created a
system where investors could outsource their
due diligence."
Stafford cites the need to focus on
measuring and judging the system's aggregate
amount of leverage and understanding the
exposures that financial institutions actually
have. "We created institutions that are too big to
fail, and that's largely because we didn't
appreciate their size and leverage," he says.
"Because of the focus on ratings, there was an
ease of avoiding critical economic analysis
throughout the whole system." As is so often
the case with large failures,
everyone—investors, ratings agencies,
borrowers, and issuers—is a little bit to blame.
About the author
Julia Hanna is associate editor of the HBS
Alumni Bulletin.
2